2003 C.I.P.A. Winners
| (back
to 2003 C.I.P.A. Winner)
Terra Payments
(formerly SureFire Commerce) Automated risk management for international payment transactions
Overview The system approves or declines every payment transaction in real time, and tracks each transaction throughout its lifecycle to improve the quality of the real-time decisions by exploiting patterns associated with legitimate versus fraudulent transactions. Objectives Previously, Terra Payments operated a risk-management workflow that was based on a less sophisticated system supplemented by a large team of fraud investigators and analysts. The company was not satisfied with the efficiency and costs associated of this workflow, but more importantly, found it difficult to identify and learn from the patterns of legitimate versus fraudulent transactions due to the manual nature of the workflow. Today, Terra Payments manages more transactions at a higher level of quality with its new system and 11 people than it did with its old system and 52 people. Summary Description To deal with the significant challenge of monitoring millions of transactions for fraud, Terra Payments employs a multi-tiered approach that benefits both the company and its clients. An automated system predicts the fraud risk of every transaction in real time. It starts by assigning a risk score to the transaction using a neural network that was trained using patterns recently identified for legitimate versus fraudulent transactions. It then calculates velocity values based on selected characteristics of the transaction. The risk score and velocity information are combined with all the other information associated with the transaction as well as the merchant past performance and fed through a rule engine that decides whether to approve or decline the transaction. This decision occurs within 500 milliseconds, prior to returning the payment authorization response to the merchant who initiated the transaction request. This system can handle multiple transactions per second. All transactions are stored in a data warehouse. Allegedly fraudulent transactions (which lead to chargebacks) are flagged in the data warehouse immediately, as they are reported. Several times each day, an automated system queries the data warehouse using a collection of rules to associate chargebacks with similar transactions that are therefore also likely to result in chargebacks. Patterns that clearly characterize a group of legitimate or fraudulent transactions are periodically fed back into the real-time neural network and rule engine to improve the quality of the decisions being made. Chargebacks may sometimes be disputed if there is evidence that the transactions they are associated with are actually legitimate, resulting in a significant financial loss recovery for our clients. An automated system compares each chargeback against a collection of rules to decide whether to dispute the chargeback. The system notifies clients on a daily basis of all new chargebacks, disputes, wins, and losses. Innovative Use of Technology Most banks and payment processing companies that serve the merchant community place the entire responsibility for transaction-level risk management on the shoulders of the merchant. Merchants that have neither the skills nor the resources to implement a sophisticated risk-management system are left with little opportunity to safely engage in high-risk transactions. By implementing a highly automated system for real-time transaction-level risk management, Terra Payments has made it possible for merchants with limited skills and resources to safely and profitably engage in high-risk transactions.
|
